OpenCode-compatible runtime provider for LiteLLM Agent Platform.

LAP already knows how to talk to an OpenCode server:

• POST /session
• POST /session/{session_id}/message
• GET /event

This service exposes those same routes. On POST /session, it starts an opencode serve process, creates a real OpenCode session, stores the session mapping, and returns the OpenCode session response. LAP can configure this service as the OpenCode runtime api_base.

OPENCODE_INFERENCE_API_KEY=sk-ant-... python3 -m opencode_api.server

If your shell has a broken global Node install, put a working Node/OpenCode path first:

PATH="$HOME/.nvm/versions/node/v20.19.5/bin:$PATH" \
OPENCODE_INFERENCE_API_KEY=sk-ant-... \
python3 -m opencode_api.server

Point OpenCode at an Anthropic-compatible gateway:

OPENCODE_INFERENCE_BASE_URL=https://litellm-rust.onrender.com \
OPENCODE_INFERENCE_API_KEY=sk-... \
python3 -m opencode_api.server

If the base URL does not end in /v1, the service appends /v1 before passing it to OpenCode as ANTHROPIC_BASE_URL.

Require bearer auth for all runtime routes except /health:

OPENCODE_API_KEY=sk-provider OPENCODE_INFERENCE_API_KEY=sk-ant-... python3 -m opencode_api.server

Set the OpenCode runtime/provider base URL to this service:

api_base = http://127.0.0.1:8088
api_key = sk-provider

The existing LAP OpenCode provider will invoke:

POST http://127.0.0.1:8088/session
POST http://127.0.0.1:8088/session/{session_id}/message
GET http://127.0.0.1:8088/event

python3 -m opencode_api.server
curl http://127.0.0.1:8088/health
curl -sS -X POST http://127.0.0.1:8088/session \
 -H 'content-type: application/json' \
 -d '{"title":"test"}'

Today this starts local opencode serve processes. The next production step is to replace SessionManager.start_opencode with sandbox creation while keeping the public routes unchanged for LAP compatibility.

This repo includes a Dockerfile for Render. Set these environment variables on the Render service:

• OPENCODE_API_KEY: bearer token LAP uses to call this provider
• OPENCODE_INFERENCE_BASE_URL: Anthropic-compatible gateway base URL, for example https://litellm-rust.onrender.com
• OPENCODE_INFERENCE_API_KEY: key sent by OpenCode to that gateway